Long-context Question Answering on NarrativeQA Fixed Chunk 2048
32.64ScoreBaseline
Evaluation Results
| Method | Links | |
|---|---|---|
| BaselineModel=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 32.64 | |
| OurModel=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 32.39 | |
| Our + ReorderModel=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 31.4 | |
| CacheBlendModel=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 29.7 | |
| EPIC (15%)Model=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 29.62 | |
| Our + ReorderModel=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 29.57 | |
| OurModel=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 28.91 | |
| No RecomputeModel=ChatGLM, Setting=Fixed Chunk (2048)2026.03 | 27.58 | |
| EPIC (15%)Model=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 27.01 | |
| CacheBlendModel=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 26.85 | |
| No RecomputeModel=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 26.39 | |
| Our + ReorderModel=Qwen, Setting=Fixed Chunk (2048)2026.03 | 22.51 | |
| CacheBlendModel=Qwen, Setting=Fixed Chunk (2048)2026.03 | 21.7 | |
| OurModel=Qwen, Setting=Fixed Chunk (2048)2026.03 | 21.1 | |
| EPIC (15%)Model=Qwen, Setting=Fixed Chunk (2048)2026.03 | 19.99 | |
| BaselineModel=Qwen, Setting=Fixed Chunk (2048)2026.03 | 16.54 | |
| BaselineModel=LLaMA, Setting=Fixed Chunk (2048)2026.03 | 16.23 | |
| No RecomputeModel=Qwen, Setting=Fixed Chunk (2048)2026.03 | 11.37 |